[Gmsh] "Perfect Match Failed" Problem with Blossom

Christophe Geuzaine cgeuzaine at ulg.ac.be
Wed Mar 13 20:33:17 CET 2013


On 13 Mar 2013, at 20:31, Christophe Geuzaine <cgeuzaine at ulg.ac.be> wrote:

> 
> Hi Johannes,
> 
> Indeed: the problem comes from the "Line in Surface" command, which is still a bit experimental.

(more precisely, we don't yet ensure that the conditions for the blossom algorithm to succeed are satisfied with these embedded lines)


> 
> For your geometry you should really define two surfaces ; that would be more "natural". Attached is an example.
> 
> Christophe
> 
> <tip.geo>
> On 13 Mar 2013, at 15:30, Johannes Reinhardt <johannes.reinhardt at unifr.ch> wrote:
> 
>> Hi,
>> 
>> sorry, that was unfortunate: I had accidentally attached the case for dist = 2.1, which works also here. For larger dists (e.g. 3.5) it fails, see new attachment.
>> 
>> It fails with all the triangulation methods, and the resulting triangulation always shows weird artifacts (see attached screenshot), but at different places depending on the triangulation method.
>> 
>> Thank you for your rapid answer
>> 
>> Johannes
>> 
>> 
>> On 03/13/2013 02:11 PM, Christophe Geuzaine wrote:
>>> 
>>> Hi Johannes,
>>> 
>>> That's weird - it seems to work fine over here (cf screenshot). I will test it on our Linux machine when I have the time.
>>> 
>>> Does it happen with all the triangulation methods (MeshAdapt, Delaunay, Frontal, Delaunay for quads)?
>>> 
>>> 
>>> Christophe
>>> <Mail Attachment.png>
>>> 
>>> On 13 Mar 2013, at 13:32, Johannes Reinhardt <johannes.reinhardt at unifr.ch> wrote:
>>> 
>>>> Hello everybody,
>>>> 
>>>> first I want to thank everybody involved in the development of gmsh. It is a really great tool and has helped me a lot in my research.
>>>> 
>>>> However I run into a probem when I try to create a 2D quad mesh with the outline of two superimposed disks embedded. However Blossom almost always fails with the message:
>>>> 
>>>> Warning : Perfect Match Failed in Quadrangulation, Applying Graph Splits
>>>> 
>>>> printed 10 times. I attached a simple geometry file that shows this problem.
>>>> 
>>>> For small distances (e.g. dist=2.1) this problem does not occur, but for larger distances (and these are the ones that I want to study), I always get this error.
>>>> 
>>>> For other geometries where I had this problem, playing a bit with the mesh element sizes helped, but I could not find good numbers for this case. I also tried both DelQuad and MeshAdapt as Meshing Algorithms, but I always get the error above.
>>>> 
>>>> I briefly looked at the Blossom Quad publication, but could not find any explanation for this behaviour. My best guess is, that the embedded curve conflicts with a perfect matching, and the extra edges mechanism fails in this case.
>>>> 
>>>> I tested both gmsh 2.6.1 compiled from source and the precompiled gmsh 2.7.0 Linux 64bit binary on Ubuntu 11.10, both show the same behaviour.
>>>> 
>>>> Has anybody an idea how to reformulate the geometry description to avoid this problem?
>>>> 
>>>> Thanks in advance and greetings
>>>> 
>>>> Johannes
>>>> <tip.geo>_______________________________________________
>>>> gmsh mailing list
>>>> gmsh at geuz.org
>>>> http://www.geuz.org/mailman/listinfo/gmsh
>>> 
>>> -- 
>>> Prof. Christophe Geuzaine
>>> University of Liege, Electrical Engineering and Computer Science 
>>> http://www.montefiore.ulg.ac.be/~geuzaine
>>> 
>>> 
>>> 
>> 
>> <Screenshot at 2013-03-13 15:22:24.png><tip.geo>_______________________________________________
>> gmsh mailing list
>> gmsh at geuz.org
>> http://www.geuz.org/mailman/listinfo/gmsh
> 
> -- 
> Prof. Christophe Geuzaine
> University of Liege, Electrical Engineering and Computer Science 
> http://www.montefiore.ulg.ac.be/~geuzaine
> 
> 
> 

-- 
Prof. Christophe Geuzaine
University of Liege, Electrical Engineering and Computer Science 
http://www.montefiore.ulg.ac.be/~geuzaine